Pigeonhole questions

It is planned for FY25. There are certain challenges associated with this that we are trying to work out. Please contact us if you have strong interest and we are happy to share and discuss further details.

Yes, Duo Chat is available upon activation of GitLab Duo License.

There’s no prescribed method for selection, but we provide guidelines in our documentation to help you decide based on your code classification. Refer to Code classification and sensitivity level.

Please refer to Code classification and sensitivity level.

GitLab Duo currently requires a separate subscription. However, for the next few months, you can use it for free with no additional charges. We are working with GitLab to streamline provisioning and explore sustainable distribution within the government. While cost recovery from agencies may be needed in the future, we aim to negotiate reasonable terms for centralised access.

All coding tools can help you write pipeline files. However, if you need direct debugging support within the GitLab UI, GitLab Duo offers that capability.

Vendors must have a public officer endorse their sign up. The sign-up form has slightly different fields for vendors, requiring the email of the endorsing officer and an attached approval letter. Additionally, all users, including vendors, must follow best practices. For more information, refer to the AI coding tools Sign up.

You can try multiple AI coding tools, but not at the same time. If you’d like to evaluate a different assistant, you can coordinate within your team to rotate access and test another tool. This helps us manage budget constraints while ensuring fair access for all interested users. If you have a specific need to use multiple assistants concurrently, let us know—we can explore possible solutions on a case-by-case basis.

We appreciate your feedback! We’re working on simplifying onboarding as part of our roadmap. Currently, most onboarding is via TechPass, so setup is straightforward—just sign in through your IDE. However, we recognise the signup-to-license activation flow can improve and are actively refining it. If you have suggestions, feel free to share them via the Telegram channel or other feedback channels!

Please contact us directly (via telegram group or our support channel) if you have such an issue, we are keen to find out how this happens as well.

It depends on your use case. To use these tools with code classified as OFFICIAL OPEN, there are no policy restrictions, and you don’t need approval.

However, if you want to use these tools for higher code classifications (e.g., RESTRICTED and above), it gets more complex.

For coding tools we officially provide, we ensure compliance through a policy clearance process—this includes verifying that prompts aren’t stored and data isn’t used for training. If you want to use a tool that is not on our list or apply it to a higher code classification, you must conduct due diligence to ensure compliance.

Yes, GSIB or COMET devices are considered GMD devices. However, they are not commonly used for software development. For AI coding tools, the primary requirement is a SEED device.

We do understand that some teams use GSIB or COMET devices for development, particularly for on-premises systems that require custom installations. As GovTech transitions to the COMET framework, we are actively exploring how AI coding tools can be supported in different environments, including GSIB or COMET devices, with respect to the following:

• Software installation: We need to ensure whether AI coding tools can be installed.
• Network access: Many AI coding tools are SaaS-based and require internet access. If network restrictions block these services, they may not work on GSIB or COMET devices.
• GCC+ Considerations: In some government environments like GCC+, SEED devices are not used, and we are assessing how to support AI coding tools in these setups.

If there is a strong use case for using AI coding tools on GSIB or COMET devices, teams are encouraged to reach out so we can explore potential solutions.

GitLab Duo uses third-party LLMs (e.g., Claude, Vertex), and data is sent to the respective LLM provider. However, we follow an AI framework to ensure responsible usage.

Yes, you can, please refer to our Code Classification Rubric using your GSIB or COMET devices.

Please follow our - User responsibilities and best practices(best-practices.md). On the platform side, SHIP-HATS has contracts in place with providers to ensure data is not stored / used for training.

This feature is coming soon. Currently, you can select a code snippet and ask Duo Chat to review it. Security scanners are also integrated into CI/CD, with real-time IDE SAST scans coming up. For more details, refer to Have GitLab Duo review your code.

Yes, you can script this in the Duo Chat prompt.

GitLab Duo provides wide coverage in security and license scanners to identify vulnerabilities or license risks. Vulnerability and license information is provided along results to help in your remediation. For copyright infringement, consult your legal team.

GitLab Duo Chat generates original code snippets based on patterns it has learned,rather than copying. A sub-processor agreement ensures that sent code is not saved or used for training.

It is available once the license is activated. Refer to Summary of GitLab Duo features for details.

For Duo and other coding tools, various IDEs are supported. Please refer to their official documentation for the full list of compatible IDEs. For example, visit GitLab documentation to know about the IDEs supported by Duo.

Compatibility depends on the specific coding assistant. Please refer to the documentation of your chosen assistant to check which cloud IDEs are supported.

Yes. GitLab Code Suggestions supports a list of languages by default, with more configurable via workflow plugins. For more details, refer to Add support for more languages.

Yes, Duo Chat can create documentation and README files.

A workflow for agentic commands is in development as part of GitLab Duo Workflow.

Yes, if proper context is provided. Code suggestions use open files and tabs as context.

Yes, but with limited capabilities. Features requiring GitLab context (e.g., issues, repos, merge requests) won’t be available.

Currently, GitLab Duo supports unmodified, API-only third-party LLMs. GitLab does not use customer AI inputs for training without consent.

Yes, you can generate database schemas with the AI coding assistant.

Yes, GitLab’s AI coding assistant can generate JUnit, Jest, and other unit tests. For more details, refer to the GitLab slides.

Yes, it does. However, the way it integrates with your workflow may vary depending on the tool you’re using.

• Interface: The AI assistant primarily works within your IDE. Some coding tools, like GitLab’s AI, integrate directly into the GitLab UI, providing a different user experience.
• Test case generation: The assistant can generate test cases based on your code and project context. You might need to fine-tune the output to match your testing requirements.
• Test reports: While AI can help describe test cases and suggest improvements, generating full test reports may require additional configuration. The ability to export or format reports depends on your specific setup.
• Running & evaluating test cases: AI can analyse test case code to check alignment with expected outcomes. Some limitations could be there in test execution and reporting.
• User involvement: As a software developer, tester, or quality engineer, you will still need to guide and validate the process to ensure the AI-generated test cases and reports are accurate and useful.

The AI coding assistant can assist with certain aspects of your CI/CD pipeline, but its capabilities depend on how you integrate it into your workflow.

• Pipeline analysis & debugging: The AI coding assistant can review your pipeline code (e.g., your .GitLab-ci.yml file) to help identify potential issues and suggest improvements.
• Troubleshoot errors: Some AI coding tools, like GitLab Duo, offer debugging support for pipeline failures.
• Code refactoring: If you open your CI/CD configuration file in your IDE, the AI assistant may suggest optimizations or refactor parts of the YAML configuration.
• Advanced debugging: For more advanced debugging or automated fixes, tools like GitLab Duo may offer additional features. Refer to the GitLab Duo demo and resources for more details.

For GitLab Dedicated, Duo features will become available once the license is activated. However, the code review feature is still in the experimental stage and will go through Beta before reaching General Availability (GA), which may take another 2–3 quarters. Most of the features demoed in the GitLab Duo launch session are already available, but some may be delayed due to the version lag between GitLab Dedicated and GitLab.com, typically trailing by one version.

No, while the AI can accelerate how you code and potentially change the team’s workflow, we are not expecting it to replace a developer.

New joiners also need to fill out the sign up form.

We will process the request and provision the necessary licences within 2–3 business days.

Currently, we don’t have an option to onboard a team. Team members need to sign up individually.

To access the coding assistant, you need a valid TechPass account.

Your access may be revoked if your account is inactive for more than a month. In that case, you need to sign up again to request access to the coding assistant.

If you have an active TechPass account and are still unable to access the tool, please reach out through our community or support channels for assistance.

When you sign up through this programme, you will receive an email confirmation from the team once your license is provisioned. After that, you can log in using your TechPass to download and install the coding assistant plugin in your IDE. We are not sure if GitHub sends a private email for individual sign-ups, but the key difference is that with this programme, the onboarding process includes guided steps and documentation available in the docs portal. If you encounter any issues, you can refer to the documentation or reach out for support.

Yes, you need to sign up again and indicate which AI tool you prefer this time. It may take 2-3 working days for us to process your request.

Yes, it is possible to use both coding tools simultaneously. However, switching between them may require adjustments. If you want to use multiple assistants concurrently, it’s recommended to write to us. Currently, while multiple assistants are allowed, the team aims to ensure fair access for all users. Since GovTech incurs costs for these tools, it’s encouraged to minimize overlapping usage to free up licenses for others. If you’re evaluating different tools and gathering insights, please share your findings with the team to help improve the overall user experience.

Refer to this link for the full features of Gitlab Duo.

Windsurf Plugin includes a few models like (Base model, Windsurf Plugin Premier & Claude 3.5 sonnet)

Yes, that’s a great suggestion! We will look into it. Additionally, Windsurf is planning to release a fully integrated JetBrains IDE in Q2, and we will evaluate this as well. Stay tuned for updates!

While ChatGPT is a more general AI tool that can handle some coding tasks, it was considered for coding assistance. However, we’d love to hear your thoughts on why ChatGPT over other tools. You can also join one of our focus groups to share your feedback.

We primarily measure coding speed and productivity through user surveys, where users report their perceived time savings when using the coding assistant. These insights are correlated with focus group interviews to ensure accuracy.

While A/B testing is the most precise method, we are currently focusing on gathering user feedback. We are also exploring potential experiments with our data science team and will inform users if participation is needed.

New joiners can apply for access in the same way as any user. They need to sign up, and we will assign the necessary licences upon submission.

Currently, we don’t have a team onboarding option. You can onboard your team by asking them to sign up.

When you sign up through this programme, you will receive an email confirmation from the team once your license is provisioned. After that, you can log in using your TechPass to download and install the coding assistant plugin in your IDE. We are not sure if GitHub sends a private email for individual sign-ups, but the key difference is that with this programme, the onboarding process includes guided steps and documentation available in the docs portal. If you encounter any issues, you can refer to the documentation or reach out for support.

While ChatGPT is a more general AI tool that can handle some of the coding tasks. In coding assistance space, we did consider ChatGPT. However, we’d love to hear your thoughts on why ChatGPT over the others. You can also join one of our focus groups to share your feedback. 

Yes, it is possible to use both coding tools simultaneously. However, switching between them may require adjustments. If you want to use multiple assistants concurrently, it’s recommended to write to us.

Currently, while multiple assistants are allowed, the team aims to ensure fair access for all users. Since GovTech incurs costs for these tools, it’s encouraged to minimize overlapping usage to free up licenses for others.

If you’re evaluating different tools and gathering insights, please share your findings with the team to help improve overall user experience.

It is possible for GitLab Duo . When you onboarded to GitLab Duo, you will have access to AI functionalities within your GitLab UI. While features like asking questions and explaining code are available, the integration for CI/CD pipelines is still developing. Currently, GitLab Duo can help with explaining reports, such as vulnerability reports, and provide assistance directly in the GitLab UI. The experience also includes support through the VS Code or IntelliJ plugin for coding tasks.

Yes, that’s a great suggestion! We will look into it. Additionally, Windsurf is planning to release a fully integrated JetBrains IDE in Q2, and we will evaluate this as well. Stay tuned for updates!

We primarily measure coding speed and productivity through user surveys, where users report their perceived time savings when using the coding assistant. These insights are correlated with focus group interviews to ensure accuracy. While A/B testing is the most precise method, we are currently focusing on gathering user feedback. We are also exploring potential experiments with our data science team and will inform users if participation is needed.

Yes, you need to sign up again and indicate which AI tool you prefer this time. It may take 2-3 working days for us to process your request.